Current gas turbine engine design practice is typically based on improvements of previ- ously successful designs. Dawes and Molinari [12] suggest that, to overcome the apparent asymptote of aero-thermal performance, a new and innovative approach to design should be adopted with 3D CFD introduced early in the design process rather than for analysis once the initial design has been created.
This chapter has reviewed the fundamental modes of boundary layer transition as well as the sources and impacts of unsteadiness within the gas turbine engine. These are two key areas that CFD solvers are currently poor at predicting accurately in the design process and as a result are two areas that require significant attention.
Improvements and advances in computational power are constantly reducing the time re- quired to solve CFD solutions. However realistic solutions to the unsteady Navier-Stokes equations are still not available as a design tool. Large Eddy Simulation (LES) CFD solutions are available, but they require a large amount of time and computational resources, which makes them an unreasonable design tool. However LES simulations still give vital insight into detailed flow characteristics, thus making them a key tool in modern CFD research.
A number of gaps exist in the the body of knowledge concerning flow characteristics in gas turbine compressors. The majority of these gaps involve understanding specific and highly complex flow phenomena in both steady and unsteady environments. Such information is of key importance to the blade and CFD code designer. This is a vital step if CFD codes are to achieve an increasingly large usage in the initial stage of an innovative design process.
This draws our attention to the flow at small distances downstream of the leading edge and concentrates research into the nature of the boundary layer under varying flow conditions in both a steady and unsteady environment.
Experimental Facilities and
Intrumentation
3.1
Experimental Facilities
The experimental investigations carried out throughout this project used a controlled dif- fusion compressor blade with a circular arc leading edge profile. The blade profile was tested inside a large-scale 2D cascade wind tunnel located in the Whittle Laboratory at the University of Cambridge, UK. The blades were many times the size of blades used in actual engines allowing for elevated Reynolds numbers and very high resolution measurements to be taken in the leading edge region.
Blades inside the two dimensional cascade had a chord, C, of 285mm and a leading edge circle of diameter 10.6mm allowing for previously unseen measurement resolution in the leading edge region of the blades. Measurements in the blade to blade plane could be taken in either a steady or unsteady flow environment with the use of rotating bar mechanism. A range of Reynolds numbers and inlet flow angles could also be covered. As a result the blade profile could be examined under a large variety of flow conditions and the resulting
data and observations could be used to compare how the blade performance alters with changing flow characteristics.
3.2
2D Cascade Description
A linear compressor cascade was used to simulate the flow through an annular compressor stator blade row. The large scale 2D cascade used in this project was situated and built in the Whittle Laboratory at the University of Cambridge and was designed with five passages using four whole blades and two half blades to complete the top and bottom passages. Each of the blades has a chord of 285mmand a pitchwise spacing of 280mmgiving the cascade a solidity of 0.99. The design inlet flow angle was 45o and the maximum achievable Reynolds number under steady operation was Rec = 400, 000 and under unsteady operation was
Rec = 300, 000. Ideally seven or more blade passages are used in a compressor cascade to
aid flow periodicity, but this was not possible due to size restrictions of the available wind tunnel exits.
The cascade was placed at the left hand exit (when facing upstream) of the Rhoden tunnel at the Whittle Laboratory - a large open circuit wind tunnel down stream of a centrifu- gal fan. Before entering the cascade the flow was controlled using a number of screens and honeycomb sections to improve uniformity before being accelerated through a 3 : 1 area contraction. Downstream of the contraction the flow passed through a turbulence grid designed to produce a turbulence intensity of 4.0% at the leading edge plain of the instru- mented blade, which, as mentioned previously, is a widely accepted value for a embedded stage of a multistage machine [65, 6].
Bleed slots were positioned at the turbulence grid and upstream of the cascade inlet plain to remove high loss boundary layer flow from the tunnel walls prior to the flow entering the cascade. End wall tip gaps were present between the blades and the cascade end walls to help control end wall flows and minimise any three dimensional flow from occurring in the adverse pressure region on the suction surface of the blades. These were shown to be highly effective using flow visualisation techniques.
Static pressure tappings were placed in the middle of each blade passage on both side walls 0.5C downstream of the trailing edge and in the leading edge plane and a pitot tube was used to provide a reference pressure downstream of the in-tunnel rotating bars in line with the blades leading edges.
axis traverse mechanism was used to perform exit area traverses of the cascade using a 3HP. Due to spatial restrictions it was not possible to place an automated traverse mechanism at the cascades inlet however an accurate, custom made, manual traverse system was devised to traverse the inlet plain through a slot milled out of the inlet section upstream of the rotating bar mechanism.
Originally the cascade was designed to be used at a fixed inlet flow angle of α1 = 45o. Modifications were made to the inlet section downstream of the turbulence grid and sup- porting frames were added to allow for a±5o variation of inlet flow angle away from the blade’s design operating point. The front of the cascade was supported on stub axles inside two thrust bearings mounted on an outrigger support system, which were aligned with the instrumented blade’s leading edge to maintain a constant distance between it and the turbulence grid. The rear of the cascade could be be raised (increase positive incidence) or lowered (increase negative incidence) using three screw jacks attached to one of the lateral supporting members at the rear of the cascade. This system, in combination with a digital inclinometer, could set the cascade to within±0.1o.
As incidence was increased or decreased the inlet flow area to the cascade must decrease or increase respectively. Two movable false floors were placed inside the inlet section to ensure the inlet area and flow angle remained consistent as the incidence of the cascade was varied. The floors were combined with an additional bell-mouth styled inlet located upstream of the turbulence grid to ensure the flow was parallel and as uniform as possible when passing through the turbulence grid. Initially due to a poorly designed inlet streamline curvature was present through the turbulence generating grid and, as a result, there was an uneven generation of loss across the grid, which seriously affected the inlet velocity distribution to the cascade. The bell-mouth inlet also had to be adjusted with any change in incidence to maintain alignment with the false floors inside the inlet. Figure 3.1(a) shows a schematic of the complete cascade and inlet mechanism, which is explained in detail in Section 3.2.2.
3.2.1 CD Stator Blade Row
The CD stator blade row was designed for a 1.5 stage research compressor at the Univer- sity of Tasmania by Hughes and Walker [43] in collaboration with Roll-Royce plc in Derby UK. The blade row was designed to incorporate features typical of modern controlled dif- fusion (CD) blading whilst maintaining the same outlet flow as the C4 blades previously used in the research compressor. Controlled diffusion blade profiles are designed around a pre-described surface velocity distribution, a method which is seen to provide significant aerodynamic benefits. A constant chord of 152.4mmwas used, which is exactly double the
(a) Schematic diagram of the 2D compressor cascade used in this research project
(b) Photograph of the 2D compressor used in this research project
Figure 3.2: Hub section, mid-span section and casing section profiles of the C4 and Cd stator blade profiles
chord of the previous C4 blade. A schematic of theC4 and CD blade profiles at hub, mid span and tip can be seen in Figure 3.2. The CD blade count was reduced from 38 to 19 during the blade row modifications. By reducing the blade count by half each of the sta- tor blades still experienced the same pitchwise flow field for a given clocking arrangement given the inlet guide vanes and rotor blades were not altered.
The stator blades were manufactured by Complete Fabrication, Whittlesford, UK from a 3D CAD model. A master blade was CNC machined from Prolab 65 modeling compound. The master blade was then measured using a DEA global coordinate measuring system and displayed an accuracy of 1.5µm. The maximum deviation from the design at mid-span was less than±0.2mmat the leading edge and less than ±0.3mmover the whole blade profile. The blade design used in the large scale cascade was taken from the mid-span profile of the CD stator blade row. Five out of the six blades in the cascade were manufactured out of modeling board with the sixth, being the instrumented blade, was manufactured on a Stereo Lythograph printer. The instrumented blade was split into three 200mm spanwise lengths due to manufacturing constraints. One section of the blade was instrumented with 54 hot- film sensors, one was left untouched for PIV measurements and one was instrumented with 194 surface static pressure tappings (Figure 3.4). Figure 3.3 shows a CAD image of the instrumented blade. The Stereo Lythograph printer with the chosen blade alignment has a documented tolerance of 32µmwhich equates to a maximum potential deviation of 0.01% chord.
Figure 3.3: 3D CAD drawing of the static pressure tapping section of the instrumented blade
Figure 3.4: Schematic drawing of the instrumented blade showing locations of hot-film sensors and static pressure tappings
Figure 3.5: Schematic diagram of the variable incidence inlet mechanism used with the 2D compressor cascade
and suction surfaces. 157 of the tappings were within the first 15% of surface length. To fit such a large number of tapping on to the blade nine hollow cavities were created under the blade’s surface. Each of the cavities had up to 15 holes drilled into it and each cavity had a single pressure line leading to a Scanivalve stepping valve and PSI pressure transducers (Section 3.5.4). The sole downfall of this technique was only one tapping could be measured in each cavity at a time and so obtaining pressure from all the cavities was time consuming.
Placing the PIV, hot-film sensors and blade static tappings on the same blade ensures that each section of the blade experiences the same inlet flow conditions and periodicity, as long as two-dimensionality is maintained in the inlet, even if the periodicity is poor (Section 3.4)
3.2.2 Variable Incidence Inlet
To allow for variations of inlet flow angle the entire cascade was rotated relative to a fixed inlet flow from the wind tunnel. A number of modification were made to the cascade itself and a new variable incidence inlet section was made (Figure 3.5 and 3.6).
As incidence is varied the effective inlet flow area changes and the inlet had to be modified to accommodate this. A system was implemented to allow the upper and lower floors of
Figure 3.6: 3D CAD drawing of the variable incidence inlet mechanism used with the 2D compressor cascade
the inlet section to be controlled and to ensure the correct inlet area was maintained and that the floors remained parallel with the upstream walls of the tunnel and inlet flow.
With an increase in incidence angle the top of the cascade moves in an upstream and down- ward direction, whilst the bottom moves in a downstream and upward direction. The false floors had to have the capacity to extend and contract to accommodate for the streamwise movement at the top and bottom of the cascade as well a be able to move vertically. The bell mouth style inlet sections upstream of the turbulence grid remained at a fixed streamwise location but had to remain aligned with the downstream inlet section and had to move vertically as the inlet flow angle changed.
Once finalised the system allowed for quick and accurate set-up times.
3.2.3 Rotating Bar Mechanism
Unsteady flow measurements were possible using a rotating bar mechanism (Figure 3.7)which consisted of two large belts holding carbon fibre bars in a spanwise orientation which shed wakes upstream of and produced periodic unsteadiness in the stator blade row. The belts traveled through a slot in the inlet of the cascade just upstream of the blade leading edges
Figure 3.7: Photograph of the rotating bars in the cascade inlet section upstream of the compressor cascade
using a series of pulleys and an electric drive motor. The bars entered through the top of the cascade and traveled parallel to the leading edge plain. An ABS electronic variable speed drive (VSD) was used to control the frequency of the drive motor to within 0.1Hz. Motor frequency could be easily converted to bar passing frequency. A trigger box and optical sensor were used to check the bar passing frequency and to trigger unsteady data acqui- sition routines. The blade passing frequency was easily altered and fixed by increasing or decreasing the motor speed using the VSD.
A derivation of stator reduced frequency is contained in Section 4.2.6 along with an de- scription of the carbon fibre bar choice and spacing options.
3.2.4 Turbulence Generating Grid
A 2D turbulence grid (Figure 3.8) was designed and placed at the exit of the wind tunnel contraction upstream of the cascade to achieve a nominal turbulence intensity of 4.0% and
Λx
Chord = 0.1 at the inlet plane of the cascade. Roache [68] shows equations 3.1, 3.2 and 3.3
can be used to determine the bar diameter and streamwise distance required to produce a designated turbulence intensity and mixing length as follows.
Figure 3.8: Photograph of the turbulence generating grid upstream of the compressor cas- cade Tu =Cx d −5 7 (3.1) Λx d =0.2 x d 1 2 (3.2)
Equating equations 3.1 and 3.2 gives
Λ Chord C d =0.2 x d 1 2 (3.3)
Solving forxanddgives
d=0.017m
x=1.14m
The nearest commercially available bar was of d = 0.019m was used and thus x was cor- rected to x = 1.26m. The resulting mixing length wasΛx = 31mmwhich corresponded to
Λx